Pathfinding algorithms for python 3.

Currently there are 7 path-finders bundled in this library, namely:

• A*
• Dijkstra
• Best-First
• Bi-directional A*
• Breadth First Search (BFS)
• Bi-directional Breadth First Search (BFS)
• Iterative Deeping A* (IDA*)
• Minimum Spanning Tree (MSP)

Dijkstra and A* take the weight of the fields on the map into account.

If you are still using python 2 take a look at the (unmaintained) python2-branch.

This library is provided by pypi, so you can just install the current stable version using pip:

pip install pathfinding

see pathfinding on pypi

For usage examples with detailed descriptions take a look at the docs folder, also take a look at the test/ folder for more examples, e.g. how to use pandas.

image_pathfinding.py in the examples/-folder provides an example how to load an image with a start and goal point. You can all it with an input and output file like this:

cd examples/
python3 image_pathfinding.py -i map.png -o foo.png

While running the pathfinding algorithm it might set values on the nodes. Depending on your path finding algorithm things like calculated distances or visited flags might be stored on them. So if you want to run the algorithm in a loop you need to clean the grid first (see Grid.cleanup). Please note that because cleanup looks at all nodes of the grid it might be an operation that can take a bit of time!

All pathfinding algorithms in this library are inheriting the Finder class. It has some common functionality that can be overwritten by the implementation of a path finding algorithm.

The normal process works like this:

1. You call find_path on one of your finder implementations.
2. init_find instantiates the open_list and resets all values and counters.
3. The main loop starts on the open_list. This list gets filled with all nodes that will be processed next (e.g. all current neighbors that are walkable). For this you need to implement check_neighbors in your own finder implementation.
4. For example in A*s implementation of check_neighbors you first want to get the next node closest from the current starting point from the open list. the next_node method in Finder does this by giving you the node with a minimum f-value from the open list, it closes it and removes it from the open_list.
5. if this node is not the end node we go on and get its neighbors by calling find_neighbors. This just calls grid.neighbors for most algorithms.
6. If none of the neighbors are the end node we want to process the neighbors to calculate their distances in process_node
7. process_node calculates the cost f from the start to the current node using the calc_cost method and the cost after calculating h from apply_heuristic.
8. finally process_node updates the open list so find_path can run check_neighbors on it in the next node in the next iteration of the main loop.

flow:

  find_path
    init_find  # (re)set global values and open list
    check_neighbors  # for every node in open list
      next_node  # closest node to start in open list
      find_neighbors  # get neighbors
      process_node  # calculate new cost for neighboring node

Because most algorithms are very similar we use inerhitance to reduce the code, however this makes it a bit harder to understand as you need to jump between the finder implementation and the finder base class, this diagram visualizes the function calls between the AStarFinder and the Finder classes as an example, this flexible aproach allows you to use inheritance to take control of processing the data and extending the algorithm. Note that this is not a classic UML sequence diagram, we just use it for visualation, feel free to suggest a nicer diagram/explaination as new issues.

sequenceDiagram
  User ->> AStarFinder: find_path(start, end, grid)
  AStarFinder ->> Finder: cleanup() [inheritance]
  Finder ->> Grid: cleanup()
  Grid ->> Grid: dirty = True
  Finder ->> AStarFinder: check_neighbors(start, end, grid, open_list) <br />[from find_path]
  AStarFinder ->> Finder: find_neighbors(graph, node) <br />[from check_neighgors]
  Finder ->> Grid: neighbors(node, ...)
  AStarFinder ->> Finder: process_node(graph, neighbor, node, end, open_list, open_value)<br />[from check_neighbors]
  Finder ->> Finder: apply_heuristic(self, node_a, node_b, ...)
  AStarFinder ->> User: return path, self.runs<br />[from find_path]

You can run the tests locally using pytest. Take a look at the test-folder

You can follow below steps to setup your virtual environment and run the tests.

# Go to repo
cd python-pathfinding

# Setup virtual env and activate it - Mac/Linux for windows use source venv/Scripts/activate
python3 -m venv venv
source venv/bin/activate

# Install test requirements
pip install -r test/requirements.txt

# Run all the tests
pytest

Please use the issue tracker to submit bug reports and feature requests. Please use merge requests as described here to add/adapt functionality.

python-pathfinding is distributed under the MIT license.

Andreas Bresser, self@andreasbresser.de

Authors and contributers are listed on github.

Inspired by Pathfinding.JS